[PATCH] preempt/hardirq.h: Fix comment about PREEMPT_ACTIVE bit location

[PATCH] preempt/hardirq.h: Fix comment about PREEMPT_ACTIVE bit location

[Srivatsa S. Bhat]

PREEMPT_ACTIVE flag is bit 27, not 28. Fix the comment.

Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
---

 include/linux/hardirq.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index bb7f309..305f23c 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -22,7 +22,7 @@
  *
  * - bits 16-25 are the hardirq count (max # of nested hardirqs: 1024)
  * - bit 26 is the NMI_MASK
- * - bit 28 is the PREEMPT_ACTIVE flag
+ * - bit 27 is the PREEMPT_ACTIVE flag
  *
  * PREEMPT_MASK: 0x000000ff
  * SOFTIRQ_MASK: 0x0000ff00

[tip:sched/urgent] sched: Fix comment about PREEMPT_ACTIVE bit location

[tip-bot for Srivatsa S. Bhat]

Commit-ID:  a7e4786b937a3ae918a7520cfdba557a80915fa7
Gitweb:     http://git.kernel.org/tip/a7e4786b937a3ae918a7520cfdba557a80915fa7
Author:     Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
AuthorDate: Sat, 21 Jul 2012 00:54:59 +0530
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 26 Jul 2012 13:54:58 +0200

sched: Fix comment about PREEMPT_ACTIVE bit location

PREEMPT_ACTIVE flag is bit 27, not 28. Fix the comment.

Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: paulmck@linux.vnet.ibm.com
Cc: josh@joshtriplett.org
Cc: rostedt@goodmis.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120720192459.6149.14821.stgit@srivatsabhat.in.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 include/linux/hardirq.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index bb7f309..305f23c 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -22,7 +22,7 @@
  *
  * - bits 16-25 are the hardirq count (max # of nested hardirqs: 1024)
  * - bit 26 is the NMI_MASK
- * - bit 28 is the PREEMPT_ACTIVE flag
+ * - bit 27 is the PREEMPT_ACTIVE flag
  *
  * PREEMPT_MASK: 0x000000ff
  * SOFTIRQ_MASK: 0x0000ff00

Re: [PATCH] preempt/hardirq.h: Fix comment about PREEMPT_ACTIVE bit location

[Steven Rostedt]

On Sat, 2012-07-21 at 00:54 +0530, Srivatsa S. Bhat wrote:
> PREEMPT_ACTIVE flag is bit 27, not 28. Fix the comment.
> 
> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
> ---
> 
>  include/linux/hardirq.h |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
> index bb7f309..305f23c 100644
> --- a/include/linux/hardirq.h
> +++ b/include/linux/hardirq.h
> @@ -22,7 +22,7 @@
>   *
>   * - bits 16-25 are the hardirq count (max # of nested hardirqs: 1024)
>   * - bit 26 is the NMI_MASK
> - * - bit 28 is the PREEMPT_ACTIVE flag
> + * - bit 27 is the PREEMPT_ACTIVE flag

Sorry but in arch/x86/include/asm/thread_info.h we have:

#define PREEMPT_ACTIVE		0x10000000

Which happens to be the 28th bit, not 27th. Although, if an arch did not
define PREEMPT_ACTIVE it would be 27th.

Perhaps the correct thing to do is to say:

  By default the following bits are defined as:
	[...]
  Different architectures may modify these.

And then add your change.

-- Steve

>   *
>   * PREEMPT_MASK: 0x000000ff
>   * SOFTIRQ_MASK: 0x0000ff00

Re: [PATCH] preempt/hardirq.h: Fix comment about PREEMPT_ACTIVE bit location

[Srivatsa S. Bhat]

On 07/31/2012 09:19 AM, Steven Rostedt wrote:
> On Sat, 2012-07-21 at 00:54 +0530, Srivatsa S. Bhat wrote:
>> PREEMPT_ACTIVE flag is bit 27, not 28. Fix the comment.
>>
>> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
>> ---
>>
>>  include/linux/hardirq.h |    2 +-
>>  1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
>> index bb7f309..305f23c 100644
>> --- a/include/linux/hardirq.h
>> +++ b/include/linux/hardirq.h
>> @@ -22,7 +22,7 @@
>>   *
>>   * - bits 16-25 are the hardirq count (max # of nested hardirqs: 1024)
>>   * - bit 26 is the NMI_MASK
>> - * - bit 28 is the PREEMPT_ACTIVE flag
>> + * - bit 27 is the PREEMPT_ACTIVE flag
> 
> Sorry but in arch/x86/include/asm/thread_info.h we have:
> 
> #define PREEMPT_ACTIVE		0x10000000
> 
> Which happens to be the 28th bit, not 27th. Although, if an arch did not
> define PREEMPT_ACTIVE it would be 27th.
> 
> Perhaps the correct thing to do is to say:
> 
>   By default the following bits are defined as:
> 	[...]
>   Different architectures may modify these.
> 
> And then add your change.
> 

Thanks a lot for your suggestions Steve! And sorry for the delay..
I'll post the updated patch (which applies on latest Linus master) as a reply
to this thread.

Regards,
Srivatsa S. Bhat

[PATCH] preempt/hardirq.h: Clarify PREEMPT_ACTIVE bit location

[Srivatsa S. Bhat]

By default, the PREEMPT_ACTIVE flag is bit 27, but different
architectures can override that. Update the comment to reflect
this fact.

Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
---

 include/linux/hardirq.h |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index 305f23c..483aa60 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -20,10 +20,15 @@
  * hardirq nesting. An arch may choose to give less than 10 bits.
  * m68k expects it to be 8.
  *
+ * By default, the bits are defined as shown below:
+ *
  * - bits 16-25 are the hardirq count (max # of nested hardirqs: 1024)
  * - bit 26 is the NMI_MASK
  * - bit 27 is the PREEMPT_ACTIVE flag
  *
+ * However, different architectures may override them. (For example,
+ * x86 defines bit 28 as the PREEMPT_ACTIVE flag).
+ *
  * PREEMPT_MASK: 0x000000ff
  * SOFTIRQ_MASK: 0x0000ff00
  * HARDIRQ_MASK: 0x03ff0000

Re: [PATCH] preempt/hardirq.h: Clarify PREEMPT_ACTIVE bit location

[Thomas Gleixner]

On Tue, 14 Aug 2012, Srivatsa S. Bhat wrote:

> 
> By default, the PREEMPT_ACTIVE flag is bit 27, but different
> architectures can override that. Update the comment to reflect
> this fact.

Is there any sensible reason why architectures need to override that?
I can't find one.

If there is none, then we should just remove all the overrides from
the arch code instead. If there is a reason, then it should be
documented.

Thanks,

	tglx

Re: [PATCH] preempt/hardirq.h: Clarify PREEMPT_ACTIVE bit location

[Steven Rostedt]

On Tue, 2012-08-14 at 12:18 +0200, Thomas Gleixner wrote:
> On Tue, 14 Aug 2012, Srivatsa S. Bhat wrote:
> 
> > 
> > By default, the PREEMPT_ACTIVE flag is bit 27, but different
> > architectures can override that. Update the comment to reflect
> > this fact.
> 
> Is there any sensible reason why architectures need to override that?
> I can't find one.
> 
> If there is none, then we should just remove all the overrides from
> the arch code instead. If there is a reason, then it should be
> documented.

IIRC, when I added the NMI bits to implement 'in_nmi()' I tried to make
this work across archs. But I found that bits in the preempt count have
been hard coded in assembly all over the place, and to clean it up
looked to be a nightmare.

It probably should be done, but I didn't have the time to do it then.
Perhaps we can revisit it.

In fact, looking at all the archs, I believe all archs override the
PREEMPT_ACTIVE bit, and not one of them uses the default (bit 27).

-- Steve